There's this odd strain of silicon valley types recently that lean hard into 'sjw' type language to promote or defend stuff that doesn't mesh with the language at all. People pushing...
There's this odd strain of silicon valley types recently that lean hard into 'sjw' type language to promote or defend stuff that doesn't mesh with the language at all. People pushing cryptocurrency as "anti-capitalist", or that guy who launched a "steroids-allowed" version of the Olympics who described steroid users like they were a marginalised group.
I'm guessing NaNoWriMo picked up this ableism narrative in the boardroom from whichever AI group they're in with and had no better ideas when it came time to make a public statement.
While I have many thoughts about using AI tools in the writing process, I will point out the obvious - this post shows how AI can smooth over inequities in editing and publishing, but notably...
While I have many thoughts about using AI tools in the writing process, I will point out the obvious - this post shows how AI can smooth over inequities in editing and publishing, but notably barely touches on actually writing. Maybe that shows how NaNoWriMo has transformed over the years into a year round service and community. Does anyone else remember when they took the website down for 9 months of the year?
It does rub me the wrong way a bit to see criticisms of use of AI as being generally classist or ableist.
NaNoWriMo's argument for this is pretty far fetched, in my opinion. From a resources perspective, if you can access an LLM, you can certainly access a notepad, digital or physical. And I'm also...
It does rub me the wrong way a bit to see criticisms of use of AI as being generally classist or ableist.
NaNoWriMo's argument for this is pretty far fetched, in my opinion. From a resources perspective, if you can access an LLM, you can certainly access a notepad, digital or physical. And I'm also not convinced that most disabilities would hinder someone from writing in some form. If an LLM is usable, so are adaptive keyboards/mice, voice to text, etc. Mental disabilities significant enough to be a hinderance would also likely be too severe to allow proofreading LLM output.
Minority voices, misunderstood voices, disabled voices, and so on have produced some of the most powerful literature because they can draw from those experiences and share them with us, opening our eyes to what we may have never seen otherwise. LLMs for writing is like using Cliff Notes to assemble a book. There may be something to it, but the artistry and honesty of storytelling is absent.
I can’t help but think of how photography was criticized by painters in the 19th century, and then how digital photography was criticized by analog photographers. I don’t think we have yet seen...
but the artistry and honesty of storytelling is absent.
I can’t help but think of how photography was criticized by painters in the 19th century, and then how digital photography was criticized by analog photographers. I don’t think we have yet seen how these tools will be used in the future, but artists are going to continue making art with the most advanced tools available to them, and it will still be art — but in a form we have yet to explore.
I think it comes down to context. There are professional photographers who would scoff at using a photoshop-esque piece of software for anything more than color correction. There are some that...
I think it comes down to context.
There are professional photographers who would scoff at using a photoshop-esque piece of software for anything more than color correction. There are some that would even look down upon that.
The new "AI-powered" adjustments on smartphone cameras make it easy for the general public to share great looking photos with Facebook friends and whatnot. But that wouldn't be viewed as a valid entry in some photography competitions, because the image is altered beyond the capabilities of a standard camera.
But that doesn't make these features a bad thing. They do have value, and they can be used to produce photos that are interesting, clearer, unique, funny, and many other outcomes if you include various "filters" that exist.
But it's a struggle for me to connect producing writing with AI augmentation. My school has a license for Grammarly, and it's helpful from time to time. But wow, sometimes it truly wants to suck the life out of my writing. (And I'm only writing essays and homework responses.) If that tool, which has been around for a long time now, takes an input of human-made writing and reduces it to something that seems cold and overly standardized, I have trouble believing any other current AI tool would be better at grammar checking without doing the same.
And that line seems much more blurry than visual art. Most works of art in a museum have a plaque next to them that state the medium(s) used. They also may have more information on the specific tools and techniques the author is known for. It's easy to ask someone about their art and get a response on how they made it, from a tools and materials perspective, even if that "tool" is AI generation.
With writing, it's much easier to be deceptive. It's possible only the author knows how the writing was assembled, and no one else will. (With current AI, it is detectable by many people, but I do suspect it will eventually get better, even if I don't consider better to mean "good.")
Not sure I fully agree with the comparison. Photography and painting are different mediums after all, photography didn't replace it. That's the main concern with AI and writing, people using it to...
Not sure I fully agree with the comparison. Photography and painting are different mediums after all, photography didn't replace it. That's the main concern with AI and writing, people using it to entirely replace the field on a professional level.
There are already news outlets using AI to write articles with mixed results, but that's still one less potential job for human writers. Meanwhile, writing is already sadly underrated within creative industries such as film and video games, and treated as secondary to the visual aspects or star power. Even though a good story and characters can make something truly legendary that sticks with people, the executives funding these projects just don't seem to care as much for that aspect.
Basically, I don't see AI itself as a threat so much as the people in charge of those industries. They just want a product that's marketable and has maximum profits, and AI is just another way to cut costs. As it stands right now AI won't cut it for longer stories such as TV series or novels, but I can see film executives wanting to just get rid of writing teams beyond the minimum necessary to review and revise the AI-written works.
That said, people WILL continue to write no matter what because we're passionate. Humans are naturally drawn to telling stories after all, and AI won't make that go away. My concern is just that AI will make it harder for truly original stories to be shared and discovered.
When professional photography became mainstream, it rapidly displaced professional painters because it was much more convenient. Photography was an existential crisis for painting, because...
Photography and painting are different mediums after all, photography didn't replace it. That's the main concern with AI and writing, people using it to entirely replace the field on a professional level.
When professional photography became mainstream, it rapidly displaced professional painters because it was much more convenient. Photography was an existential crisis for painting, because capturing reality was the main value of painting at the time.
Fair point, though I'd argue against that being the main value of painting. Artists in the early 1800's weren't just capturing reality like portraits and landscapes, but also scenes from...
Fair point, though I'd argue against that being the main value of painting. Artists in the early 1800's weren't just capturing reality like portraits and landscapes, but also scenes from mythology, literature and history—scenes that couldn't be captured in photographs. Though you're right that it did kill off portrait painting, which was one of the primary ways for painters to make money.
That gives me an interesting thought, though: while painting is no longer primarily used to capture reality, I think that the rise of photography helped encourage painters to focus on more creative and abstract styles rather than photorealism. It also provided references for painters to use, or gave them inspiration by showing them subjects and places they could never see with their own eyes.
So now I wonder if AI could have some comparable impact on writing, though I can't imagine what that might look like.
I think that's one of my bigger frustrations with AI to be honest. If it were being used, for example, to give people more time for creative endeavors by automating work duties that can be done...
I think that's one of my bigger frustrations with AI to be honest. If it were being used, for example, to give people more time for creative endeavors by automating work duties that can be done safely or to ensure that basic tasks are tracked so someone can spend more time on individual conversations with an employee or a client or something. Like that is useful automation to me. Grammar checks make sense, something that could do a first pass of editing would be great. But doing, and advertising doing, things that suck up the creative energy and experience and take away the joyful things rather than the drudgery, or just use it as an excuse to add more work to existing employees, that's sort of the worst.
People imagine AI tools being used to write essays for you, but they can also be used for more mundane writing tasks like checking for grammar errors. It seems reasonable to believe that people...
People imagine AI tools being used to write essays for you, but they can also be used for more mundane writing tasks like checking for grammar errors. It seems reasonable to believe that people such as immigrants who aren't as fluent in English might find these tools especially useful.
Properly used, it won't look like obvious AI slop. It will just look a bit better written. I expect the real harm will come from accusations of using AI tools when writing is somehow "too good." Better to say that they're okay, as long as used in ways that the reader wouldn't notice.
For better or for worse I'm pretty sure like 60% of the text on arxiv these days is written in chatGPT. LLMs bailing out grad students from having to learn English.
It seems reasonable to believe that people such as immigrants who aren't as fluent in English might find these tools especially useful.
For better or for worse I'm pretty sure like 60% of the text on arxiv these days is written in chatGPT. LLMs bailing out grad students from having to learn English.
I'm a native English speaker, but I've still done experiments with using LLMs for papers that I've worked on recently. I don't actually use it to write the text for me directly. (I am of the...
I'm a native English speaker, but I've still done experiments with using LLMs for papers that I've worked on recently.
I don't actually use it to write the text for me directly. (I am of the opinion that if I, as a human, am going to ask another human to read it, then I, as a human, should be the one to write it.)
But what I've experimented with is feeding the text of a draft into a LLM and quizing it on how it interprets my prose, or how it would summarize it. If it gets anything wrong, I'll follow up and ask it what led to that conclusion. Sometimes this reveals ambiguous wording in something I've written.
Along those lines, I've also sometimes asked an LLM to compare a paragraph before and after my editing and tell me which one seems better. I've found that they can be surprisingly good at picking up on shades of nuance between two similar words and helping me think through which one might be closer to my intent.
And when asking an LLM how I might trim an overly wordy passage, I've been occasionally surprised when they pop out with a wording or even word that I hadn't thought of but that is perfect.
Asking their opinion on comma placement, capitalization, and hyphenation has also been useful.
Basically, I think LLMs still make fairly lousy writers, but surprisingly good editors when work-shopping a draft interactively.
I think this shows how people are still figuring out how to use these tools. I don’t know if this is good history, but it’s like the electric motor was just invented. At one time, apparently there...
I think this shows how people are still figuring out how to use these tools.
I don’t know if this is good history, but it’s like the electric motor was just invented. At one time, apparently there were companies that sold electric motors with various attachments for different tasks. I imagine it sort of worked, but was awkward? It’s a long way from there to selling an electric razor or a vacuum cleaner. Other parts need to be invented and often years of product evolution to get something good.
Similarly, I expect that there will be good writing tools and they won’t work like today’s writing tools. They probably won’t look like the first products companies are building with AI, either. Most of those will fail, but some will probably be useful.
I appreciate what you are saying, but for a person who could use an adaptive technology, for example a gaze driven keyboard, to eek out a sentence or two per minute, to drive something that could...
I appreciate what you are saying, but for a person who could use an adaptive technology, for example a gaze driven keyboard, to eek out a sentence or two per minute, to drive something that could accurately express their intent in fuller language at a speed that matches 'abled conversation ' would be immensely empowering.
Adaptive tech can enable, but it cannot always equalize. There is much room for improvement and augmentation of the existing tools in that space.
NaNoWriMo really decided that their reputational losses from allegedly having a moderator grooming teens on their youth forums and ignoring it until after the FBI was notified left them with room...
NaNoWriMo really decided that their reputational losses from allegedly having a moderator grooming teens on their youth forums and ignoring it until after the FBI was notified left them with room to go downwards.
There's this odd strain of silicon valley types recently that lean hard into 'sjw' type language to promote or defend stuff that doesn't mesh with the language at all. People pushing cryptocurrency as "anti-capitalist", or that guy who launched a "steroids-allowed" version of the Olympics who described steroid users like they were a marginalised group.
I'm guessing NaNoWriMo picked up this ableism narrative in the boardroom from whichever AI group they're in with and had no better ideas when it came time to make a public statement.
Related: NaNoWriMo is in disarray after organizers defend Al writing tools [The Verge]
While I have many thoughts about using AI tools in the writing process, I will point out the obvious - this post shows how AI can smooth over inequities in editing and publishing, but notably barely touches on actually writing. Maybe that shows how NaNoWriMo has transformed over the years into a year round service and community. Does anyone else remember when they took the website down for 9 months of the year?
It does rub me the wrong way a bit to see criticisms of use of AI as being generally classist or ableist.
NaNoWriMo's argument for this is pretty far fetched, in my opinion. From a resources perspective, if you can access an LLM, you can certainly access a notepad, digital or physical. And I'm also not convinced that most disabilities would hinder someone from writing in some form. If an LLM is usable, so are adaptive keyboards/mice, voice to text, etc. Mental disabilities significant enough to be a hinderance would also likely be too severe to allow proofreading LLM output.
Minority voices, misunderstood voices, disabled voices, and so on have produced some of the most powerful literature because they can draw from those experiences and share them with us, opening our eyes to what we may have never seen otherwise. LLMs for writing is like using Cliff Notes to assemble a book. There may be something to it, but the artistry and honesty of storytelling is absent.
I can’t help but think of how photography was criticized by painters in the 19th century, and then how digital photography was criticized by analog photographers. I don’t think we have yet seen how these tools will be used in the future, but artists are going to continue making art with the most advanced tools available to them, and it will still be art — but in a form we have yet to explore.
I think it comes down to context.
There are professional photographers who would scoff at using a photoshop-esque piece of software for anything more than color correction. There are some that would even look down upon that.
The new "AI-powered" adjustments on smartphone cameras make it easy for the general public to share great looking photos with Facebook friends and whatnot. But that wouldn't be viewed as a valid entry in some photography competitions, because the image is altered beyond the capabilities of a standard camera.
But that doesn't make these features a bad thing. They do have value, and they can be used to produce photos that are interesting, clearer, unique, funny, and many other outcomes if you include various "filters" that exist.
But it's a struggle for me to connect producing writing with AI augmentation. My school has a license for Grammarly, and it's helpful from time to time. But wow, sometimes it truly wants to suck the life out of my writing. (And I'm only writing essays and homework responses.) If that tool, which has been around for a long time now, takes an input of human-made writing and reduces it to something that seems cold and overly standardized, I have trouble believing any other current AI tool would be better at grammar checking without doing the same.
And that line seems much more blurry than visual art. Most works of art in a museum have a plaque next to them that state the medium(s) used. They also may have more information on the specific tools and techniques the author is known for. It's easy to ask someone about their art and get a response on how they made it, from a tools and materials perspective, even if that "tool" is AI generation.
With writing, it's much easier to be deceptive. It's possible only the author knows how the writing was assembled, and no one else will. (With current AI, it is detectable by many people, but I do suspect it will eventually get better, even if I don't consider better to mean "good.")
(This response is directed to @skybrian as well.)
Not sure I fully agree with the comparison. Photography and painting are different mediums after all, photography didn't replace it. That's the main concern with AI and writing, people using it to entirely replace the field on a professional level.
There are already news outlets using AI to write articles with mixed results, but that's still one less potential job for human writers. Meanwhile, writing is already sadly underrated within creative industries such as film and video games, and treated as secondary to the visual aspects or star power. Even though a good story and characters can make something truly legendary that sticks with people, the executives funding these projects just don't seem to care as much for that aspect.
Basically, I don't see AI itself as a threat so much as the people in charge of those industries. They just want a product that's marketable and has maximum profits, and AI is just another way to cut costs. As it stands right now AI won't cut it for longer stories such as TV series or novels, but I can see film executives wanting to just get rid of writing teams beyond the minimum necessary to review and revise the AI-written works.
That said, people WILL continue to write no matter what because we're passionate. Humans are naturally drawn to telling stories after all, and AI won't make that go away. My concern is just that AI will make it harder for truly original stories to be shared and discovered.
When professional photography became mainstream, it rapidly displaced professional painters because it was much more convenient. Photography was an existential crisis for painting, because capturing reality was the main value of painting at the time.
Fair point, though I'd argue against that being the main value of painting. Artists in the early 1800's weren't just capturing reality like portraits and landscapes, but also scenes from mythology, literature and history—scenes that couldn't be captured in photographs. Though you're right that it did kill off portrait painting, which was one of the primary ways for painters to make money.
That gives me an interesting thought, though: while painting is no longer primarily used to capture reality, I think that the rise of photography helped encourage painters to focus on more creative and abstract styles rather than photorealism. It also provided references for painters to use, or gave them inspiration by showing them subjects and places they could never see with their own eyes.
So now I wonder if AI could have some comparable impact on writing, though I can't imagine what that might look like.
I think that's one of my bigger frustrations with AI to be honest. If it were being used, for example, to give people more time for creative endeavors by automating work duties that can be done safely or to ensure that basic tasks are tracked so someone can spend more time on individual conversations with an employee or a client or something. Like that is useful automation to me. Grammar checks make sense, something that could do a first pass of editing would be great. But doing, and advertising doing, things that suck up the creative energy and experience and take away the joyful things rather than the drudgery, or just use it as an excuse to add more work to existing employees, that's sort of the worst.
People imagine AI tools being used to write essays for you, but they can also be used for more mundane writing tasks like checking for grammar errors. It seems reasonable to believe that people such as immigrants who aren't as fluent in English might find these tools especially useful.
Properly used, it won't look like obvious AI slop. It will just look a bit better written. I expect the real harm will come from accusations of using AI tools when writing is somehow "too good." Better to say that they're okay, as long as used in ways that the reader wouldn't notice.
For better or for worse I'm pretty sure like 60% of the text on arxiv these days is written in chatGPT. LLMs bailing out grad students from having to learn English.
I'm a native English speaker, but I've still done experiments with using LLMs for papers that I've worked on recently.
I don't actually use it to write the text for me directly. (I am of the opinion that if I, as a human, am going to ask another human to read it, then I, as a human, should be the one to write it.)
But what I've experimented with is feeding the text of a draft into a LLM and quizing it on how it interprets my prose, or how it would summarize it. If it gets anything wrong, I'll follow up and ask it what led to that conclusion. Sometimes this reveals ambiguous wording in something I've written.
Along those lines, I've also sometimes asked an LLM to compare a paragraph before and after my editing and tell me which one seems better. I've found that they can be surprisingly good at picking up on shades of nuance between two similar words and helping me think through which one might be closer to my intent.
And when asking an LLM how I might trim an overly wordy passage, I've been occasionally surprised when they pop out with a wording or even word that I hadn't thought of but that is perfect.
Asking their opinion on comma placement, capitalization, and hyphenation has also been useful.
Basically, I think LLMs still make fairly lousy writers, but surprisingly good editors when work-shopping a draft interactively.
I think this shows how people are still figuring out how to use these tools.
I don’t know if this is good history, but it’s like the electric motor was just invented. At one time, apparently there were companies that sold electric motors with various attachments for different tasks. I imagine it sort of worked, but was awkward? It’s a long way from there to selling an electric razor or a vacuum cleaner. Other parts need to be invented and often years of product evolution to get something good.
Similarly, I expect that there will be good writing tools and they won’t work like today’s writing tools. They probably won’t look like the first products companies are building with AI, either. Most of those will fail, but some will probably be useful.
For an example of mundane utility while writing papers, here’s how Terence Tao uses AI.
I appreciate what you are saying, but for a person who could use an adaptive technology, for example a gaze driven keyboard, to eek out a sentence or two per minute, to drive something that could accurately express their intent in fuller language at a speed that matches 'abled conversation ' would be immensely empowering.
Adaptive tech can enable, but it cannot always equalize. There is much room for improvement and augmentation of the existing tools in that space.
NaNoWriMo really decided that their reputational losses from allegedly having a moderator grooming teens on their youth forums and ignoring it until after the FBI was notified left them with room to go downwards.